IMO, it makes sense when your strategic advantage is in your community and the service/convenience of hosting the app for people. If your advantage is in your algorithms, I don't think it makes sense, unless you plan to make your money from consulting/enterprise/hosting.

And even then, look at how many general computer repair shops are competing against Cononical for "professional Ubuntu support".

Joel's article from 2002 is highly relevant here http://www.joelonsoftware.com/articles/StrategyLetterV.html

In 1999, Eric Raymond wrote about this in chapter 6 of his book The Magic Cauldron: "Reasons for Closing Source": http://www.catb.org/~esr/writings/magic-cauldron/magic-cauld...

I would be extremely interested in that analysis, but sadly we don't have the data to make any interesting conclusions about big products like that. Our preliminary analysis suggests that for products even several times larger than ours, it still makes a lot of sense to at least seriously consider it.

Parent brings a good point: How do you define your strategic advantage, when you're nullifying the one of your code? It's hard to come to investors if you say "Anyone can make exactly our software within minutes". Do you have Android-like trademarks which would prevent others from doing it?

note that you can open source part of your code, and keep other parts private -- many companies are quite comfortable exposing the user-facing parts of their system without explaining all of the backend